maemo.org - Talk

maemo.org - Talk (https://talk.maemo.org/index.php)
-   Applications (https://talk.maemo.org/forumdisplay.php?f=41)
-   -   [DEVEL] Saera: Siri clone for Maemo5, Harmattan and Sailfish OS (https://talk.maemo.org/showthread.php?t=84753)

Amboss 2015-04-21 17:12

Re: [DEVEL] Saera: Siri clone for Maemo5 and Harmattan
 
@taixzo:
By what you explained I should think of Saera more like my car's voice control (controlling Phone commands, Navigation, music...) then Siri or Google Now, right? Maybe this comparison helps others as well scaling their expectations ;)

taixzo 2015-05-03 05:12

Re: [DEVEL] Saera: Siri clone for Maemo5 and Harmattan
 
Quote:

Originally Posted by Amboss (Post 1468021)
@taixzo:
By what you explained I should think of Saera more like my car's voice control (controlling Phone commands, Navigation, music...) then Siri or Google Now, right? Maybe this comparison helps others as well scaling their expectations ;)

That is correct, at least as far as voice control goes. Text input should have a bit more flexibility, and will gain more down the road as I finish my NLP library.

Also, I have discovered a rather annoying bug: alarms set by Saera do not show up in the Clock app as repeating, but go off every day. I need to learn more about how alarms work to fix this.

Edit: fixed in Github, I will upload a new build later.

taixzo 2015-05-25 04:38

Re: [DEVEL] Saera: Siri clone for Maemo5 and Harmattan
 
New feature added for Sailfish: Saera starts listening when you raise the phone to your ear (based on proximity sensor). This improves two things: it makes it possible to use without looking at all, and it makes the voice recognition better in higher-noise environments. Code is on Github, I will build a new RPM later this week when I have access to my laptop with Sailfish SDK again.

taixzo 2015-05-26 03:07

Re: [DEVEL] Saera: Siri clone for Maemo5 and Harmattan
 
Another feature, and one I'm rather proud of: Saera can now scan the music on your device, and build voice models for the names of all the songs it finds (well, currently just any mp3s - I haven't looked into how tags on other audio files work). This means that you can now just say "Play (name of song)" for pretty much any song on your Jolla and Saera will recognize and play it. I made a blog post about it because I spent a while searching the internet for how to program it, so if someone is looking for it in the future there will be instructions.

javispedro 2015-05-26 08:58

Re: [DEVEL] Saera: Siri clone for Maemo5 and Harmattan
 
Quote:

Originally Posted by taixzo (Post 1471525)
well, currently just any mp3s - I haven't looked into how tags on other audio files work

You need to use Tracker for that -- unfortunately I'm not sure how the Python bindings look like.

Astaoth 2015-05-26 14:26

Re: [DEVEL] Saera: Siri clone for Maemo5 and Harmattan
 
Hi,
Does Saera supports only English or also other languages ?

taixzo 2015-05-26 14:38

Re: [DEVEL] Saera: Siri clone for Maemo5 and Harmattan
 
Quote:

Originally Posted by Astaoth (Post 1471552)
Hi,
Does Saera supports only English or also other languages ?

Only English at the moment, as a lot of the grammar rules are highly language-specific.

taixzo 2015-05-27 19:09

Re: [DEVEL] Saera: Siri clone for Maemo5 and Harmattan
 
New feature: Saera now parses contacts as well and allows you to call someone by name, e.g. say "Call John Smith" or "Call Bob". A caveat is that the speech model is trained on English phonemes, meaning that if you have a lot of contacts with names from other cultures Saera may not be able to recognize them.

taixzo 2015-05-28 20:52

Re: [DEVEL] Saera: Siri clone for Maemo5 and Harmattan
 
New feature in heavy development: turn-by-turn navigation.

szopin 2015-05-28 21:07

Re: [DEVEL] Saera: Siri clone for Maemo5 and Harmattan
 
Probably a bad idea, but... BillyBass vs Saera? When billy shouts 2+ emails received, trigger saera to read out all their subjects? (possibly even asking if you want the body read out loud too?) Will need to check if Saera understands what Billy says

TTS in question: https://openrepos.net/content/kimmoli/billy-bass
Would be nice to combine efforts (if possible), maybe get the 'read out aloud' notifications code from billy incorporated?


All times are GMT. The time now is 22:04.

vBulletin® Version 3.8.8